✦   ✦   ✦

breaking news

News, read through The Primal Race
← 全部评论 · all commentary

AI 出海,谁在付账Who Pays for AI Going Abroad

其他 CNBC ↗ 2026-11-02 § 链接
新闻摘要OpenAI 在阿联酋签署能源、投资和政府领域多项协议,CEO 奥特曼在阿布扎比会见王储哈立德。此前有报道称 OpenAI 正寻求在阿联酋投资者共同领投下融资 200 亿美元。
The newsOpenAI signed deals across energy, investment and government in the UAE, with CEO Sam Altman meeting Abu Dhabi Crown Prince Khaled bin Mohamed bin Zayed Al Nahyan. OpenAI is reportedly seeking $20 billion in a funding round co-led by UAE investors.
大模型进政府,先改的不是效率,是解释权
When models enter government, the first thing rewritten is interpretation.

Altman 在阿布扎比签下的不是三份合同,是一整套认知入口。能源、投资、政府,三个词摆在一起,说的是同一件事:让 OpenAI 的模型成为这个国家读政策、做投资判断、分配资源的默认通道。

最该盯住的是那条政府协议——把政府的政策和法规翻译成机器可读的形式。政策一旦机器可读,谁来读、按什么逻辑读、读出来给谁用,就不再是公民和议会的公共事务。它变成了模型输出。这是第三章说的认知入口:不是篡改事实,是抢先定义事实以什么形态存在。法规的语义边界被写进模型的那一刻,解释权就换了手。

Mubadala、Inception42 和 OpenAI 合作做投资分析,同一套模型既读监管规则又参与资本配置。规则的解释者和规则的受益者坐在同一张桌子上。这在任何国家都值得追问,在主权基金就是政府本身的地方,追问的空间更小。

Altman 说这是“让 AI 的好处惠及所有人”,说阿布扎比在树立“大胆的全球标准”。标准这个词在这里要拆开看:谁定的标准,谁就是以后所有人的默认设置。主权基金的钱进了 200 亿的融资轮,模型进了政府的工作流,回报是双向的,只是不写在同一条新闻里。

报道里没有一个普通人的名字。没有写这套系统上线后谁的申请被模型筛掉、谁的签证被算法判定、哪个行业的从业者在“政府效率”的名义下失去议价能力。ADNOC 和 G42 部署前沿技术,能源行业的人力成本怎么变,报道没写。没写不等于没发生,只是还没轮到被记录。

中东被写成 AI 新枢纽,冲突被一句“考验了吸引力”带过。资本流动的叙事里,地区的人从来只是背景板。真正的问题不是 OpenAI 在中东站稳了脚,是它站稳的方式:用政府的规则当训练数据,用政府的钱当融资,再用政府的背书去说服下一个政府。

What Altman signed in Abu Dhabi isn't three contracts. It's a set of cognitive entry points. Energy, investment, government — three words that say one thing: OpenAI's models become the default channel through which this state reads its policies, makes investment calls, allocates resources.

The government deal is the one to watch — translating policies and regulations into machine-readable form. Once policy is machine-readable, who reads it, by what logic, and for whom stops being a public matter between citizens and parliament. It becomes model output. This is the cognitive entry point from Chapter 3: not falsifying facts, but getting there first to define what form facts take. The moment the semantic boundary of a regulation is written into the model, interpretation has changed hands.

Mubadala, Inception42 and OpenAI applying AI to investment analysis means the same model reads the regulatory rules and participates in capital allocation. The interpreter of rules and the beneficiary of rules sit at one table. That deserves scrutiny anywhere; where the sovereign fund is the government, there's less room to ask.

Altman calls it making AI's benefits reach everyone, calls Abu Dhabi a bold global standard. Unpack the word standard: whoever sets it becomes everyone's default. Sovereign money goes into the $20 billion round, the model goes into the government workflow. The return runs both ways, just not in the same paragraph.

No ordinary person appears in this report. Nothing about whose application the model filters out, whose visa the algorithm rejects, which industry loses bargaining power under the banner of government efficiency. ADNOC and G42 deploy frontier technology — what happens to the human cost in energy, the report doesn't say. Not written doesn't mean not happening; it means not yet recorded.

The Middle East is framed as an AI hub, and conflict gets one clause about testing its appeal. In the narrative of capital flows, the people of the region are scenery. The real question isn't whether OpenAI has a foothold in the Middle East. It's how it got one: government rules as training data, government money as funding, government endorsement to convince the next government.